CN108961747A - A kind of urban road traffic state information extracting method under incomplete bayonet data qualification - Google Patents

Abstract

This patent discloses a method for extracting urban road traffic status information under the condition of incomplete bayonet data, which specifically includes the following steps: Step 1, preprocessing of incomplete data; Step 2, complementing missing data; Step 3 , to extract road traffic state information, the present invention is applicable to the extraction of urban road traffic state information, based on incomplete bayonet data, use optimized depth-first traversal to find lost track points, and then use Lagrangian polynomial interpolation method The time information of the lost track point is obtained, and then the road traffic state information can be extracted.

Description

A kind of urban road traffic status information extraction under the condition of incomplete checkpoint data method

technical field

The invention relates to the field of traffic information, in particular to a method for extracting urban traffic state information under the condition of incomplete bayonet data.

Background technique

With the improvement of my country's comprehensive national strength and people's living standards, the number of motor vehicles is increasing at a rate of 10% to 20% every year. In order to strengthen the supervision of motor vehicles and the construction of urban roads, the number of bayonet equipment has also been greatly increased. promote. The road checkpoint monitoring system can monitor the checkpoint around the clock and record all types of passing vehicles, and the recorded information is rich, including the checkpoint number, license plate number, passing time and other information. The use of bayonet data can extract and analyze urban road traffic status information. However, due to the technical level of bayonet monitoring system equipment and the number of bayonets installed by bayonet monitoring equipment, it is inevitable that information such as loss of recorded information and missing recording points will occur. The phenomenon of incomplete records makes it difficult to extract urban traffic status information under the condition of incomplete bayonet data. Therefore, it is necessary to study the extraction method of urban road traffic status information under the condition of incomplete checkpoint data.

At present, there are many studies on the extraction of road traffic state information. The patent application number 201510938922.1 - "A method for extracting characteristic parameters of traffic operation state based on big data" is based on GPS data and mobile phone signaling data for vehicle speed and road Flow parameters are extracted in a way that does not apply to bayonet data. There are still a large number of technical solutions for extracting traffic information based on checkpoint data in the prior art, but the usual practice is to discard the data when the checkpoint data is incomplete, such as the patent application number 201510225291. Calculation method of road real-time traffic speed based on large-scale checkpoint passing data" Although it is based on checkpoint data to extract road section speed information, it is not applicable to missing checkpoint data.

Contents of the invention

The invention proposes a method for extracting traffic state information when part of the bayonet data is lost. By supplementing the effective bayonet information, the recorded position information, time information and vehicle information are relatively complete, and the city Extraction of road traffic status information.

In order to solve the above technical problems, the technical solutions provided by this patent include:

A method for extracting urban road traffic status information under the condition of incomplete bayonet data is provided, said method comprising the following steps: Step 1, preprocessing of incomplete data

1.1 Vehicle trip division

Data preprocessing includes selecting checkpoint data of all recording points in the road network from the database, and eliminating abnormal data and duplicate data; classifying this part of data according to the license plate number, and assigning a number to each vehicle; chronologically Arrange the data of each vehicle; finally, divide the vehicle itinerary, the division method is as follows:

(1) According to the average time required for the normal driving of each road section in the road network, set the road section transit time set T={T ₁ , T ₂ ,...,T _V }, and the maximum value T _u (u∈V) in the set is taken as threshold.

(2) In the above data sorted by time, if the time difference between every two adjacent data points of each vehicle is greater than the threshold, it is considered that the vehicle has stayed between these two record points, and the data needs to be transferred from the two records The points are disconnected, so the data of the former and the former of the two recording points are regarded as trip A, and the latter and the following recording points are regarded as trip B. As shown in Figure 2, if the time spent by a certain vehicle between the bayonets numbered 187 and 024 is greater than the threshold, it is considered that the vehicle has stopped between the two bayonets, so its driving trajectory will be It is disconnected and divided into trip A and trip B. However, if there is still a situation in the itinerary B where the time difference between adjacent data is greater than the threshold, then the itinerary B may continue to be divided into itineraries.

(3) Eliminate the itinerary with only one record, and assign numbers to the remaining itineraries to record the division of the itinerary.

1.2 Calculation of some evaluation indicators

Preliminary calculation of the flow and speed of each road section, as the two evaluation indicators required for subsequent entropy weight method calculation, the specific calculation method is as follows:

(1) Take t as the period (t can be 5min, 10min, etc.) to divide a day into n periods, traverse the data of each period, extract the bayonet number k of each piece of data, the license plate number m, and pass the vehicle Time s and other information, calculate the number of vehicles passing through each checkpoint in each direction in each time period, and then obtain the flow information q _dj of each period of each road section, where d is the number of the road section, and j is the jth time period .

(2) Still take t as the cycle to calculate the speed v _dj of the road section, read the bayonet number k of the data, the license plate number m, the passing time s and other information according to the itinerary, and calculate the time difference between every two adjacent data: Δs=s _{i +1} -s _i , and then divide the length of the corresponding section by the time difference to get the speed, the calculation formula is as follows:

Where v _di is a speed value of the road section d connected by the bayonet number k of the i-th data and the bayonet number k+1 of the i+1 data, Δs is the time of the i+1 data and the i The time difference of the data, d _d is the length of road segment d.

Calculate the average speed of each road section for the data of each cycle, which is the speed of the road section in this cycle, the formula is as follows:

Among them, v _dj is the average speed of road section d in the jth time period, ∑v _dj is the sum of all speed values v _di of road section d in the jth time period, and h is v _di in the jth time period the number of . Store the road section number d, time section j, and road section average speed v _dj into the database, and then the speed of each road section at each time period can be obtained.

(3) Due to the lack of original bayonet data, the traffic and speed of the road section are incomplete. If the flow or speed of a road section is missing in a certain period of time, the flow or speed of the adjacent time period of this road section can be used as an approximate replacement.

Step 2, complete the missing data

If the position information displayed by the data of a certain trip can appear in each adjacent checkpoint in the road network in sequence, that is, there is no breakpoint in the trajectory information, then the data of this trip is not missing, and can be directly used to extract road traffic status information. For itinerary A shown in Figure 2, there is no missing data in the itinerary; if the location information displayed by the data of a certain itinerary cannot be connected sequentially in the road network, that is, there are breakpoints in its trajectory information, then the data of this itinerary is missing, and it needs to be corrected. It can only be used after the data of each breakpoint is supplemented completely. As for stroke B shown in Figure 2, data is missing at bayonet number 026, and the trajectory positions displayed by the stroke data are 028 to 186. Traversing the data of each trip, extracting its location information, and judging whether there are breakpoints in its trajectory, extracting the missing trip data for supplement.

2.1 Calculate the location of the missing point

For missing data, it is necessary to use optimized depth-first traversal to calculate the trajectory of the missing part. The specific steps are as follows:

(1) Use the depth-first traversal method to find out all possible paths of the starting point and the ending point corresponding to the missing part of the data, and perform initial screening on all possible paths to find out the shortest path and the difference between the number of checkpoints passed by the shortest path and the shortest path is less than or equal to 3, let the set of possible paths for the missing part of a trip be R={r ₁ , r ₂ ,...,r _n }, where r _i represents the i-th path in the set of paths. where n is the total number of possible paths that qualify.

(2) Select path length, path average speed, number of road sections, and path average traffic as evaluation indicators. The path length is the sum of the lengths of the sections that each possible path passes through, expressed as L={l ₁ , l ₂ ,...,l _n } ^T in a set, where l _i (i∈n) represents the i-th possible path The sum of the lengths of each road section passed, that is, l _i =∑d _d ; the average speed of the path is the average speed of each road section passed by each possible path in the corresponding time period, expressed as V={v ₁ , v ₂ ,…,v _n } ^T , where v _i (i∈n) represents the average speed v _dj of each road section passed by the i-th possible path, namely The number of road sections is the number of road sections that each possible path passes through, expressed as a set C={c ₁ ,c ₂ ,...,c _n } ^T , where c _i (i∈n) represents the i-th possible path passed through The number of road sections; the average flow rate of the path is the average flow rate of each road section passed by each possible path in the corresponding time period, which is expressed as Q={q ₁ , q ₂ ,...,q _n } ^T , where q _i (i∈n) represents the average value of the flow q _dj of each road section passed by the i-th possible path, namely The entropy weight method is used to determine the weight of each evaluation index. First, the initial matrix is obtained:

(3) Among the above four evaluation indicators, the average speed of the road section and the flow rate of the road section are the high-quality indicators, and the length of the road section and the number of road sections are the low-quality indicators. The different indicators should have the same trend, so the reciprocal method is used to transform the low-quality indicators into High-quality indicators, the transformed matrix is:

(4) Normalize the Y matrix, that is, use the ratio of the element yi _j of each column vector in the Y matrix to the sum of all elements of the vector as the corresponding element of the matrix Z obtained by normalization, after normalization The matrix of is:

(5) Determine the entropy weight H(x _j ) of each evaluation index, (j=1,2,3,4), the specific formula is:

where k is the adjustment coefficient, z _ij is the standardized value of the jth evaluation indicator of the i-th evaluation unit, that is, the element in the i-th row and j-th column in the Z matrix.

(6) Convert the entropy value of the evaluation index into a weight value to obtain the weight value of each evaluation index. The specific formula is:

Among them, d _j is the weight of the evaluation index in the jth column, and 0≤d _j ≤1, m is the number of evaluation indicators, that is, m=4.

(7) Determine the comprehensive evaluation value of the entropy weight of each evaluation index, multiply the weight of each index with its corresponding index standardized value and then sum, the formula is:

Among them, U _i is the comprehensive evaluation value of the entropy weight of the i-th possible path; the path with the largest comprehensive evaluation value of the entropy weight is selected as the trajectory of the missing part of the data.

2.2 Time to calculate missing points

Use the Lagrangian polynomial interpolation method to calculate the time information of the track points where the data is missing, and then get the traffic flow, speed, and average traffic density of the road. The specific steps are as follows:

(1) Let the previous bayonet recording point of the missing portion of the trajectory data be point O, and the subsequent bayonet recording point of the missing portion of the trajectory data be point D. If there is data before point O, let the previous bayonet record of point O be Point A is point A, and if there is no data before point O, point A is equal to point O. Let the distance from point O to point A be x ₀ , the distance from point D to point A be x ₁ , and the set of distances from each track point to point A in the data missing part is X={x ₁ ,x ₂ ,…,x _d }, where x _u is the distance from the uth point to point A. The time points at point O and point D are converted into time stamps and recorded as y ₀ and y ₁ respectively.

(2) Using the Lagrangian interpolation polynomial to calculate the time value of the track points in the missing part of each data, the set of time values of each track point P={p ₁ ,p ₂ ,…,p _d } can be obtained, where p _u is the time value of the uth point, and the specific formula of the Lagrangian interpolation polynomial is:

where 1≤u≤d, and x _u ∈X.

(3) Convert the time value in the set P into the time format in the original data, and add the bayonet number, license plate number, passing time and other information of the corresponding track point to the vehicle itinerary data to get the itinerary information make it complete. Integrate the completed data and non-missing data for the next step.

Step 3: Extract road traffic status information

(1) Taking t as the time period (t can be 5min, 10min, etc.), divide the time of a day into n time periods, traverse the data of each period, extract the bayonet number k and license plate number m of each piece of data, and pass Calculate the number of vehicles in each direction passing through each checkpoint in each time period, and then obtain the flow rate q _dj of each road section in each period, where d is the road section number, and j is the jth time period . Store the link number d, time period j, and link traffic q _dj into the database, and then the traffic of each link at each time period can be obtained.

(2) Still take t as the cycle to calculate the speed v _dj of the road section, read the bayonet number k of the data, the license plate number m, the passing time s and other information according to the itinerary, and calculate the time difference between every two adjacent data: Δs=s _{i +1} -s _i , and then divide the length of the corresponding section by the time difference to get the speed, the calculation formula is as follows:

Where v _di is a speed value of the road section d connected by the bayonet number k of the i-th data and the bayonet number k+1 of the i+1 data, Δs is the time of the i+1 data and the i The time difference of the data, d _d is the length of road segment d.

Calculate the average speed of each road section for the data of each cycle, which is the speed of the road section in this cycle, the formula is as follows:

Among them, v _dj is the average speed of road section d in the jth time period, ∑v _dj is the sum of all speed values v _di of road section d in the jth time period, and h is v _di in the jth time period the number of . Store the road section number d, time section j, and road section average speed v _dj into the database, and then the speed of each road section at each time period can be obtained.

(3) Calculate the average density of the road section with t as the cycle, and divide the flow of the road section in each cycle by the average speed of the corresponding road section in turn to obtain the average density of the road section in this cycle. The calculation formula is: Store the road segment number d, time segment j, and road segment density k _dj into the database, and the density of each road segment at each time period can be obtained.

The present invention is applicable to the extraction of urban road traffic state information, based on incomplete bayonet data, using optimized depth-first traversal to obtain lost track points, and then using Lagrangian polynomial interpolation method to find the time of lost track points Information, and then the road traffic status information can be extracted.

Description of drawings

Fig. 1 is the extraction flowchart of urban road traffic state information that the present invention proposes;

Fig. 2 is a schematic diagram of the division of vehicle journeys proposed by the present invention;

Fig. 3 is a schematic diagram of bayonet position in the example of the present invention;

Fig. 4 is the speed change diagram of 13 road sections in the example of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with accompanying drawings and examples.

The method for extracting urban road traffic state information under the condition of incomplete bayonet data proposed by the present invention mainly includes: data preprocessing, estimating road section flow and speed, extracting missing data, calculating the position of missing track points, and extracting road traffic state There are five small steps for information, and the process is shown in Figure 1, specifically:

Step 1, preprocessing of missing data

1.1 Vehicle trip division

Extract the bayonet data of all record points in the road network from the database according to the bayonet number information, and remove abnormal data such as records where the license plate number cannot be recognized, and then delete the data with the same bayonet number, license plate number, and passing time ;Arrange the data after removing abnormal data and duplicate data according to the license plate number, and assign a number to each vehicle; and time-arranged data; finally, the vehicle trips are divided according to the following steps:

(1), according to the average time required for the normal driving of each road section in the road network, set the road section transit time set T = {T ₁ , T ₂ ,..., T _V }, where V needs to be determined according to the number of road sections in the road network, T _u (u∈V) is the average travel time of the uth road segment. Take the largest value in the set It is used as a threshold to determine whether the vehicle trip needs to be divided.

(2), in the data that has been sorted by vehicle number and time, select two adjacent data of each vehicle in turn and find the time difference. If the time difference is greater than the threshold, it is considered that the vehicle has stayed between these two record points , the data needs to be disconnected from the two record points to divide it into two trips, so the former and the previous data of the two record points are regarded as trip A, and the latter and the subsequent record points are regarded as For itinerary B. However, if there is still a situation in the itinerary B where the time difference between adjacent data is greater than the threshold, then the itinerary B may continue to be divided into itineraries.

(3) The data that has been divided into trips may have only one record for a certain trip. This kind of trip needs to be eliminated, and the remaining trips should be assigned incremental numbers to record the trip division.

This example intercepts a vehicle record with the license plate number "Anhui CXXXXX" and the time information "2017-06-01" as an example to illustrate the data format after the vehicle itinerary is divided. The information contained in the data is as follows:

Table 1 "Anhui CXXXXX" vehicle trip division table

Bayonet No. License plate number passing time itinerary number 035 Anhui CXXXXX 2017-06-0111:28:15 1 533 Anhui CXXXXX 2017-06-0111:30:47 1 054 Anhui CXXXXX 2017-06-0120:34:54 2 026 Anhui CXXXXX 2017-06-0120:39:54 2 343 Anhui CXXXXX 2017-06-0121:35:20 3 024 Anhui CXXXXX 2017-06-0121:39:27 3 187 Anhui CXXXXX 2017-06-0121:42:10 3

Among them, 1, 2, and 3 in the itinerary number respectively represent the first section of the journey, the second section of the journey, and the third section of the itinerary of the vehicle. Each section of the journey includes the number information of the bayonet monitoring equipment and the information recording time.

1.2 Calculation of some evaluation indicators

Because the follow-up calculation needs to use the flow and speed information of the road section, the flow and speed of each road section are initially calculated. Since the checkpoint data is incomplete, the calculated flow and speed information must be missing, and the missing part needs to be approximated. deal with. The specific calculation method is as follows:

(1) Divide a day into 144 time periods with a period of 10 minutes, use the time information in each piece of data as the basis for judgment, traverse the data of each 10-minute period, and extract the bayonet number of each piece of data k, license plate number m, passing time s and other information, each time a vehicle passes through a checkpoint, a passing record will be generated, so when calculating the traffic, it is only necessary to count the consecutive passing through certain two adjacent checkpoints within each 10-minute period The number of pieces of data is the flow q _dj of the road section where the two adjacent checkpoints are located. If a vehicle passes through the first checkpoint of a road section, its recorded time is in the previous 10-minute period, and when it passes through the second checkpoint, its recorded time is in the next 10-minute period, then it is included in the previous 10-minute period traffic. Calculated according to this method, the flow information of each 10-minute cycle of each road section can be obtained, and finally the road section number d, time section j, and flow q _dj are stored in the database for subsequent calculation.

(2), still calculate the speed of the road section with a cycle of 10 minutes, read the bayonet number k of the data, the license plate number m, the passing time s and other information according to the itinerary, calculate the time difference Δs between two adjacent data and calculate it in seconds Stored as a unit, the time difference of the last piece of data in each trip is recorded as 0, and then the length of the corresponding road section of two adjacent data is divided by the time difference to obtain a speed of the corresponding road section in the 10-minute period corresponding to the passing time value v _di , and then calculate the average value of the speed data of each 10-minute cycle of each road segment, and v _dj is the speed of the road segment in the 10-minute cycle. The speed information v _dj is stored in the database according to the corresponding road section and the corresponding time period j, so as to prepare for the subsequent calculation.

(3) Due to the lack of original checkpoint data, the traffic and speed of the road section are incomplete. It may happen that the data of the flow or speed of a certain road section in a certain 10-minute period is missing, then the flow or speed of the adjacent period of this road section can be used as an approximate substitute.

This example intercepts the speed of the road section with the road section number 1 when the time information is "2017-06-0103:40:00" to "2017-06-0104:10:00", to illustrate the data format and calculation process of the speed of the road section , the data is shown in the table below:

Table 2 Section speed table

section number section speed time period start time period end 1 9.750 2017-06-0103:40:00 2017-06-0103:50:00 1 9.243 2017-06-0103:50:00 2017-06-0104:00:00 1 9.180 2017-06-0104:00:00 2017-06-0104:10:00

Wherein, the speed unit of the road section is "m/s", the checkpoint number of the uplink direction of the road section 1 is 175, and the checkpoint number of the downlink direction is 442. In order to calculate the speed of the first time period in this example, it is necessary to find out all the vehicles that pass through the bayonet numbers 175 and 442 successively and pass through the bayonet 175 so that the time information is at "2017-06-0103:40:00" For the data between "2017-06-0103:50:00", divide the length of road section 1 by the time difference of each group of data, and then calculate the average value, that is, road section 1 in "2017-06-0103:40: 00" to "2017-06-0103:50:00" the average speed in the period of 10 minutes.

Step 2, complete the missing data

Traversing the data of each trip, if the bayonet numbers of every two adjacent data of a certain trip are equal to the bayonet numbers of the uplink direction and the downlink direction of a road section, then the information of this trip is complete, and the bayonet position is recorded according to the time Sequentially displayed on the map can be connected along the road to form a complete track. This type of data has no missing information and can be directly used to extract road state information; If there is a lack of class data, extract the starting point number and end point number of the missing part, and prepare for the calculation of the bayonet number information of each trajectory point in the missing part.

2.1, calculate the position of the missing point

For missing data, it is necessary to use optimized depth-first traversal to calculate the trajectory of the missing part. The specific steps are as follows:

(1), given the bayonet numbers of the starting point and the ending point, the corresponding location information can be obtained, and the depth-first traversal method is used to find all possible paths of the starting point and the ending point by using the road network connection relationship, and all possible paths are calculated. For the first screening, find out the shortest path and all paths with the difference of the number of checkpoints passed by the shortest path and the number of checkpoints that are less than or equal to 3. Since the urban road network selected in this example is in a square format, the path screening process can be simplified and the number of checkpoints that pass through the shortest path can be selected. and all paths whose number of bayonets differs by less than or equal to 3 from the path with the least number of bayonets. This example takes the missing part of the data in itinerary 3 in Table 1 as an example to illustrate the calculation of the missing point position. The data between bayonet number 343 and 024 is missing, and the set of possible paths obtained by using depth-first traversal and initial screening of its possible paths is R={r ₁ , r ₂ ,...,r _n }, where r _i represents the i-th path in the set of paths. Where n is the total number of possible paths that meet the conditions. In this example, four possible paths are obtained after screening, so n=4. Among them, r ₁ represents the first possible path, and the numbers of the bayonets passed are: 343, 344, 345, 024; r ₂ represents the second possible path, and the numbers of the bayonets passed are: 343, 344, 028 , 024; r ₃ represents the third possible path, and the numbers of the bayonets passed are: 343, 026, 028, 024; r ₄ represents the fourth possible path, and the numbers of the bayonets passed are: 343,023 , 186, 026, 028, 024, the position corresponding to each bayonet number is shown in Figure 3.

(2), select path length, path average speed, number of road sections, and path average traffic as evaluation indicators to determine the most idealized road section in the above possible path set as the trajectory of the missing part, where the path length is the path passed by each possible path The sum of the lengths of each road section is represented as L={l ₁ , l ₂ ,...,l _n } ^T by a set, and the set of path lengths calculated in this example is L={2147, 2169, 1950, 2725} ^T ; The average path speed is the average speed of each road section passed by each possible path in the corresponding time period, expressed as V={v ₁ , v ₂ ,...,v _n } ^T in a set. In this example, the set of average speed of each possible path in the period from "2017-06-0103:30:00" to "2017-06-0103:40:00" is calculated as V={7.095, 6.044, 5.786, 6.104 } ^T ; the number of road sections is the number of road sections passed by each possible path, which is represented by a set as C={c ₁ , c ₂ ,...,c _n } ^T , the set of possible path road sections obtained in this example is C= {3, 3, 3, 5} ^T ; the average flow rate of the path is the average flow rate of each road section passed by each possible path in the corresponding time period, expressed as Q={q ₁ , q ₂ ,...,q _n } ^T . In this example, the average traffic set of each possible path in the period from "2017-06-0103:30:00" to "2017-06-0103:40:00" is calculated as Q={104, 60, 51, 50 } ^T ; Use the entropy weight method to determine the weight of each evaluation index, first obtain the initial matrix, and bring the data into:

(3) Among the four evaluation indexes of path length, path average speed, number of road sections, and path average flow, the average road speed and road flow are high-quality indicators, that is, when the value is high, it is beneficial to vehicle traffic. The length of the road section and the number of road sections are low-quality indicators, that is, when the value is low, it is conducive to vehicle traffic. Different indicators should have the same trend, so the reciprocal method is used to convert the low-quality indicators into high-quality indicators, that is, to calculate the reciprocal of the low-quality indicators, and the value of the high-quality indicators remains unchanged, and write the Y matrix according to the corresponding position. The transformed matrix is:

(4), the Y matrix is normalized, that is, the ratio of the element y _ij of each column vector in the Y matrix to the sum of all elements of the vector is used as the corresponding element of the matrix Z obtained by normalization, and this example is as follows The normalized matrix obtained by the formula is:

(5), determine the entropy weight H(x _j ) of each evaluation index, (j=1,2,3,4), the specific formula is:

where k is the adjustment coefficient, z _ij is the standardized value of the jth evaluation indicator of the i-th evaluation unit, that is, the element in the i-th row and j-th column in the Z matrix. Specifically in this example, the corresponding elements of the Z matrix obtained in (4) are brought into the above formula to obtain:

H(x ₁ )=0.3543, H(x ₂ )=0.3554, H(x ₃ )=0.3510, H(x ₄ )=0.3530,

(6), convert the entropy value of the evaluation index into the weight value d _j , (j=1,2,3,4), that is, calculate the weight value of each evaluation index, and the specific formula is:

Among them, d _j is the weight of the evaluation index in the jth column, and 0≤d _j ≤1, m is the number of evaluation indicators, that is, m=4. Specifically, in this example, H(x _j ), (j=1,2,3,4) obtained in (5) is brought into the above formula to obtain:

d ₁ =0.2487, d ₂ =0.2482, d ₃ =0.2510, d ₄ =0.2530

(7) Determine the comprehensive evaluation value of the entropy weight of each evaluation index, multiply the weight of each index with its corresponding index standardized value and then sum, the formula is:

Among them, U _i is the comprehensive evaluation value of the entropy weight of the i-th possible path; the path with the largest comprehensive evaluation value of the entropy weight is selected as the trajectory of the missing part of the data. Collectively to this example, put d _j in (6), (j=1,2,3,4), and the corresponding elements in the Z matrix into the above formula:

U ₁ =0.3029, U ₂ =0.2504, U ₃ =0.2467, U ₄ =0.2009

Since the evaluation function value U ₁ is the largest, the first possible path in the possible path set is selected as the trajectory of the missing part, and the position information of each bayonet can be known from the number information of each bayonet in the first possible path.

2.2, Calculate the time of missing points

After the location information of the missing point is obtained, the data is still incomplete, and its time information needs to be obtained before it can be used. In this example, the Lagrangian polynomial interpolation method is used to calculate the time information of the track points where the data is missing, and to prepare for the subsequent calculation of road traffic, road speed, and road average traffic density. The specific steps are as follows:

(1), it is assumed that the previous bayonet recording point of the missing portion of the trajectory data is point O. In this example, the starting point of the possible path is point O, that is, the bayonet numbered 343 is point O. The last bayonet recording point of the track data missing part is point D. In this example, the end point of the possible path is point D, that is, the bayonet numbered 024 is point D. There is no data before point O in this itinerary, so point A is equal to point O. Let the distance from point O to point A be x ₀ , that is, x ₀ =0, the distance from point D to point A be x ₁ , that is, x ₁ =2147m, and the set of distances from each track point to point A in the data missing part is J={j ₁ ,j ₂ ,...,j _d }, where j _u (u∈[1,d]) is the distance from the uth point to point A. The J={653,1182} obtained in this example are respectively the distances from each checkpoint in the calculated path to point A. Convert the time formats of points O and D into seconds, and record them as y ₀ and y ₁ respectively. In this example, the time points at points O and D are converted into time stamps, and y ₀ =1496324120s and y ₁ =1496324367s are obtained.

(2), using the Lagrangian interpolation polynomial to calculate the time value of the track points in the missing part of each data, the set of time stamps of each track point P={p ₁ ,p ₂ ,…,p _d } can be obtained, where p _u (u∈[1, d]) is the time stamp of the uth point, and the specific formula of the Lagrangian interpolation polynomial is:

where 1≤u≤d, and x _u ∈X. Put the values obtained in this example into the formula according to the requirements:

P={1496324195,1496324255}

(3), convert the time stamp in the set P to the time format in the original data, after conversion: p ₁ is 2017-06-0121:36:35, p ₂ is 2017-06-0121:37:35 . And add the bayonet number, license plate number, passing time and other information of the corresponding track point to the vehicle travel data to complete the travel information. In this example, the itinerary data is supplemented as shown in Table 3. Finally, integrate the supplementary and complete itinerary data with the non-missing itinerary data for use in the next step.

Table 3 Supplementary itinerary 3 information table

Bayonet No. License plate number passing time itinerary number 343 Anhui CXXXXX 2017-06-0121:35:20 3 344 Anhui CXXXXX 2017-06-0121:36:35 3 345 Anhui CXXXXX 2017-06-0121:37:35 3 024 Anhui CXXXXX 2017-06-0121:39:27 3 187 Anhui CXXXXX 2017-06-0121:42:10 3

Step 3: Extract road traffic status information

After the missing data is completed and the data without missing is integrated, the flow, speed and density of the road section can be calculated. The calculation method is as follows:

(1) Divide a day into 144 time periods with a period of 10 minutes, use the time information in each piece of data as the basis for judgment, traverse the data of each 10-minute period, and extract the bayonet number of each piece of data , license plate number, passing time and other information, every time a vehicle passes through a checkpoint, a passing record will be generated, so when calculating the traffic, it is only necessary to count the records of data passing through two adjacent checkpoints in each 10-minute period. The number is the flow rate of the road section where the two adjacent checkpoints are located. If a vehicle passes through the first checkpoint of a road section, its recorded time is in the previous 10-minute period, and when it passes through the second checkpoint, its recorded time is in the next 10-minute period, then it is included in the previous 10-minute period traffic. According to this calculation method, the flow information of each road section in each 10-minute cycle can be obtained, and finally the road section number, time period, and flow are stored in the database, and the traffic flow of each road section at each time period can be obtained.

(2), still calculate the speed of the road section with a cycle of 10 minutes, read the bayonet number, license plate number, passing time and other information of the data according to the itinerary, calculate the time difference between each two adjacent data and store it in seconds, The time difference of the last piece of data in each trip is recorded as 0, and then the length of the corresponding road section of the two adjacent data is divided by the time difference to obtain a speed value of the corresponding road section in the 10-minute period corresponding to the passing time, and then for each The average speed data of each 10-minute period of a road section is the speed of the road section within the 10-minute period. The speed information is stored in the database according to the corresponding road section and corresponding time period, and the speed of each road section at each time period can be obtained. In this example, 13 road sections with different road grades are selected to draw the speed change map within a day, as shown in Figure 4.

(3) Calculate the average density of the road section in a period of 10 minutes, and divide the flow of the road section in each 10-minute period by the corresponding average speed of the road section to obtain the average density of the road section for this period. Store the road section number, time period, and road section density in the database to get the density of each road section at each time period.

The above is the implementation of specific cases of the present invention and the principle of its application. If the present invention is changed under the conception of the present invention, but its function is still beyond the spirit covered by the description and accompanying drawings, it still belongs to the protection of the present invention. scope.

Claims (3)

1. the urban road traffic state information extracting method under a kind of incomplete bayonet data qualification, it is characterised in that including such as Lower step:

Step 1, the pretreatment of incomplete data

1.1 vehicle travels divide

The pretreatment of data includes recording the bayonet data put, and rejecting abnormalities data from all in selection road network in database And repeated data；This partial data is classified according to license plate number, and distributes and numbers to each car；In chronological order to each The data of vehicle are arranged；Finally, dividing to vehicle travel, division methods are as follows:

(1) section transit time set T={ T is arranged by required average time according to section each in road network normally travel₁, T₂..., T_V, the maximum of T in set_u(u ∈ V) is used as threshold value.

(2) in above-mentioned data according to time sequence, if every two adjacent data time differences of each car are greater than threshold value, recognize There is stop between the two record points for vehicle, needs data from the separated of the two record points, so by the two Record puts the former and its pervious data and is considered as stroke A, and the latter and its later record point are considered as stroke B, if in=stroke B still The case where having the adjacent data time difference to be greater than threshold value, can then continue stroke division to stroke B；

(3) stroke of only one record is rejected, and to remaining stroke distribution number to record stroke dividing condition；

1.2 calculating section evaluation indexes

Each link flow of primary Calculation and speed, two evaluation indexes needed for being calculated as subsequent entropy assessment, specific calculating side Method is as follows:

(1) one day time is divided into n period using t as the period, the data in each period is traversed, extract each item number According to bayonet number k, license plate number m, cross the information such as vehicle time s, calculate all directions in each time cycle by each bayonet Vehicle number, the flow information q in you can get it each section each period_dj, wherein d is section number, and j is j-th of time cycle；

(2) section speed v still is calculated by the period of t_dj, bayonet number k, the license plate number m of data are read by stroke, spend the vehicle time The information such as s calculate the time difference per adjacent two data: Δ s=s_i+1-s_i, then can divided by the time difference with corresponding road section length Speed is obtained, calculation formula is as follows:

Wherein v_diBy the bayonet number k of the i-th data and one of the bayonet number k+1 of the i+1 data section d connected Velocity amplitude, Δ s are the difference of the time of i+1 data and the time of i data, d_dFor the length of section d；

Seeking the speed average in its each section to the data in each period, i.e. the section speed in the period, formula are as follows thus:

Wherein, v_djFor the speed average of the section d in j-th of time cycle, ∑ v_djFor the section d of j-th time cycle All velocity amplitude v_diSum, h is v in j-th time cycle_diNumber.By section number d, period j, road average-speed v_djIt is stored in database, each section can be obtained in the speed of each period；

(3) when the flow or speed in certain section have missing in certain period, then close using the flow of this section adjacent time interval or speed Like substitution.

Step 2, completion missing data

If if location information shown by the data of certain stroke could be successively present in each adjacent bayonet of road network, i.e. its rail Breakpoint is not present in mark information, then this trip data are without missing, directly using extracting road traffic state information；If certain stroke Location information shown by data cannot sequentially link in road network, i.e., there are breakpoints for its trace information, then this trip data have Missing just can be used after then supplementing completely the data of each of which breakpoint；The data for traversing each stroke extract its location information, And its track is judged with the presence or absence of breakpoint, there are the run-length datas of shortage of data to be supplemented for extraction；

2.1 calculate the position of missing point

Data in the presence of missing are calculated with the track of its lack part using the depth-first traversal of optimization, specific steps are such as Under:

(1) all possible paths of the corresponding beginning and end in shortage of data part are found out using the method for depth-first traversal, And all possible paths are screened for the first time, find out wherein shortest path and differ small with the bayonet number that shortest path is passed through In all paths for being equal to 3, if the collection of the possible path of certain stroke lack part is combined into R={ r₁, r₂..., r_n, wherein r_iTable Show the i-th paths in the set in path.Wherein n is the sum of qualified possible path；

(2) select path length, path average speed, section quantity and path average flow rate are as evaluation index；Path length The length in each section that degree is passed through for every possible path be L={ l with set expression₁, l₂..., l_n}^T, wherein l_i(i∈ N) length and i.e. l in each section that i-th possible path is passed through are indicated_i=∑ d_d；Path average speed can energy circuit for every The speed average of each section that diameter is passed through during that corresponding time period, is V={ v with set expression₁, v₂..., v_n}^T, wherein v_i (i ∈ n) indicates the average speed v in each section that i-th possible path is passed through_djAverage value, i.e., Section quantity It is C={ c with set expression by the quantity in the section that every possible path passes through₁, c₂..., c_n}^T, wherein c_i(i ∈ n) is indicated The quantity in the section that i-th possible path is passed through；Each section that path average flow rate is passed through by every possible path is right The flow average value in the period is answered, is Q={ q with set expression₁, q₂..., q_n}^T, wherein q_i(i ∈ n) indicates i-th possibility The flow q in each section that path is passed through_djAverage value, i.e.,It is carried out really with weight of the entropy assessment to each evaluation index It is fixed, initial matrix is obtained first:

(3) road average-speed and link flow are high excellent index, road section length and section quantity in aforementioned four evaluation index For low excellent index, high excellent index is converted for low excellent index using counting backward technique, the matrix after conversion are as follows:

(4) Y matrix is normalized, i.e., with the element y of each column vector in Y matrix_ijWith the vector all elements The corresponding element for the matrix Z that the ratio of sum is obtained as normalization, the matrix after normalization are as follows:

(5) the entropy weight H (x of each evaluation index is determined_j), (j=1,2,3,4), specific formula are as follows:

Wherein k is adjustment factor,z_ijFor the standardized value of j-th of evaluation index of i-th of evaluation unit, i.e., in Z matrix I-th row, the element of jth column；

(6) weighted value is converted by the entropy of evaluation index, the weight of each evaluation index, specific formula can be obtained are as follows:

Wherein, d_jFor the weight of the evaluation index of jth column, and 0≤d_j≤ 1,M is the number of evaluation index, i.e. m= 4。

(7) weight of each index, is distinguished corresponding criterion by the entropy weight comprehensive evaluation value for determining each evaluation index Value is summed after being multiplied, formula are as follows:

Wherein, U_iFor the entropy weight comprehensive evaluation value of i-th possible path；The maximum path of entropy weight comprehensive evaluation value is chosen as number According to the track of lack part.

2.2 calculate the time of missing point

Using the temporal information of shortage of data partial traces point at Lagrangian type polynomial interpolation calculating, then you can get it Link flow, section speed, road-section average vehicle density, the specific steps of which are as follows:

(1) the previous bayonet record for setting track data lack part is put as O point, the latter bayonet of track data lack part Record point is D point, and enabling the previous bayonet of O point record point if having data before O point is A point, if no data A before O point Point etc. and O point；The distance for enabling O point to A point is x₀, the distance of D point to A point is x₁, each tracing point of shortage of data part to A point The collection of distance be combined into X={ x₁,x₂,…,x_d, wherein x_uThe distance of A point is arrived for u-th point；The time point of O point and D point is turned Timestamp is turned to, and is denoted as y respectively₀, y₁；

(2) time value that each shortage of data partial traces point is calculated using Lagrangian type interpolation polynomial, can be obtained each rail Set P={ the p of the time value of mark point₁,p₂,…,p_d, wherein p_uFor u-th point of time value, Lagrangian type interpolation polynomial The specific formula of formula are as follows:

Wherein 1≤u≤d, and x_u∈X；

(3) time format in initial data is converted by the time value in set P, and the bayonet of corresponding tracing point is compiled Number, license plate number, the excessively information such as vehicle time are added in this vehicle travel data can be complete by this trip information supplement.It will supplement Data after complete and without missing Data Integration, to use in next step.

Step 3 extracts road traffic state information

(1) one day time was divided into n period using t as the time cycle, the data in each period is traversed, extracted each Bayonet number k, the license plate number m of data cross the information such as vehicle time s, calculate in each time cycle by each of each bayonet The vehicle number in direction, the flow q in you can get it each section each period_dj, wherein d is section number, and j is j-th of time cycle. By section number d, period j, link flow q_djIt is stored in database, each section can be obtained in the flow of each period.

(2) section speed v still is calculated by the period of t_dj, bayonet number k, the license plate number m of data are read by stroke, spend the vehicle time The information such as s calculate the time difference per adjacent two data: Δ s=s_i+1-s_i, then can divided by the time difference with corresponding road section length Speed is obtained, calculation formula is as follows:

Wherein v_diBy the bayonet number k of the i-th data and one of the bayonet number k+1 of the i+1 data section d connected Velocity amplitude, Δ s are the difference of the time of i+1 data and the time of i data, d_dFor the length of section d.

Seeking the speed average in its each section to the data in each period, i.e. the section speed in the period, formula are as follows thus:

Wherein, v_djFor the speed average of the section d in j-th of time cycle, ∑ v_djFor the section d of j-th time cycle All velocity amplitude v_diSum, h is v in j-th time cycle_diNumber.By section number d, period j, road average-speed v_djIt is stored in database, each section can be obtained in the speed of each period；

(3) road-section average density is calculated by the period of t, successively using the link flow in each period divided by corresponding road-section average The road-section average density in this period, calculation formula can be obtained in speed are as follows:By section number d, period j, section is close Spend k_djIt is stored in database, each section can be obtained in the density of each period.

2. the urban road traffic state information extraction side under a kind of incomplete bayonet data qualification according to claim 1 Method, which is characterized in that in 1.2 (3) of the step one, if the flow for the section period acquired or speed have missing, The flow or speed approximate substitution that adjacent time interval can be used, if the front and back adjacent time interval of specially numerical value loss period has number According to then with the average value of front and back adjacent time interval come the value of this loss period of approximate substitution；If with the presence of multiple continuous time numerical value Missing, then centre is substituted with the average value of front and back adjacent time interval, and all there are the values of numerical value loss period.

3. the urban road traffic state information extraction side under a kind of incomplete bayonet data qualification according to claim 1 Method, it is characterised in that in the 2.1 of the step two, the depth-first traversal of optimization is using depth-first traversal and entropy assessment In conjunction with optimal way, find possible miss path using depth-first traversal, the superiority and inferiority in path then judged with entropy assessment, To obtain the result of depth-first traversal of the most possible track for lack part as optimization.